Implications of Reactions Between SO2 and Basaltic Glasses for the Mineralogy of Planetary Crusts

Basalts are ubiquitous in volcanic systems on several planetary bodies, including the Earth, Mars, Venus, and Jupiter's moon Io, and are commonly associated with sulfur dioxide (SO2) degassing. We present the results of an experimental study of reactions between SO2 and basaltic glasses. We examined Fe‐free basalt, and Fe‐bearing tholeiitic and alkali basalts with a range of Fe3+/Fetotal (0.05 to 0.79) that encompass the oxygen fugacities proposed for most terrestrial planetary bodies. Tholeiitic and alkali basalts were exposed to SO2 at 600, 700, and 800 °C for 1 hr and 24 hr. Surface coatings formed on the reacted basalts; these contain CaSO4, MgSO4, Na2SO4, Na2Ca(SO4)2, Fe2O3, Fe3O4, Fe‐Ti‐(Al)‐oxides, and TiO2. Additionally, the SO2‐basalt reaction drives nucleation of crystalline phases in the substrate to form pyroxenes and possible Fe‐oxides. A silica‐rich layer forms between the substrate and sulfate coatings. More oxidized basalts may readily react with SO2 to form coatings dominated by large Ca‐sulfate and oxide grains. On less oxidized basalts (NNO−1.5 to NNO−5), reactions with SO2 will form thin, fine‐grained aggregates of sulfates; such materials are less readily detected by spectroscopy and spectrometry techniques. In contrast, in very reduced basalts (lower than NNO−5), typical of the Moon and Mercury, SO2 is typically a negligible component in the magmatic gas, and sulfides are more likely.


Introduction
Basalt is the most common volcanic rock on the terrestrial planets and is typically associated with gaseous sulfur dioxide (SO 2 ; Basaltic Volcanism Study Project, 1981;Taylor & McLennan, 2009). On Earth, basalt makes up over half of the volcanic crust, including most of the volcanoes on the seafloor, and significant quantities of SO 2 are derived from both explosive and effusive basaltic volcanoes (Carn et al., 2016). On Mars, basaltic volcanism provided the primary source of S to the surface through both SO 2 emissions, sulfides, and sulfates; these materials were redistributed as abundant sulfate minerals on the surface (Franz et al., 2018;Gaillard & Scaillet, 2009;King & McLennan, 2010;McSween et al., 2006). On Venus, basalts are the dominant volcanic rock type (McSween et al., 2006;Nimmo & McKenzie, 1998). Thermodynamic and experimental data suggest that SO 2 interacted with the surface of basalt deposits on Venus in the past (Berger et al., 2019;Fegley & Prinn, 1989;Zolotov, 2018). Basalt dominates the surface of Io, moon of Jupiter (Blaney et al., 1995;Carr, 1986), and the magmas are sufficiently oxidizing to produce SO which 2 has been directly measured in the atmosphere and eruptive plumes (Doute, 2002;Glaze, 1999;Kumar, 1985;Pearl et al., 1979;Vandaele et al., 2017). On the Moon, basalt is also a dominant rock type (Haskin & Warren, 1991); however, SO is likely negligible because oxygen fugacity ( 2 fO ) of the volcanic gas and 2 the magmatic source were too low (Renggli et al., 2017). Finally, Mercury has vast flood basalts (Denevi et al., 2013) and basaltic pyroclastic deposits (Kerber et al., 2011), and surface sulfur concentrations of up to 4 wt. % (Nittler et al., 2011;Starr et al., 2012;Weider et al., 2012).
Sulfur dioxide degassing from explosive and effusive eruptions provides the opportunity for SO 2 reactions with glassy particles to scavenge SO 2 and form sulfates (e.g., Ayris et al., 2013;Barone et al., 2016;Delmelle et al., 2018;Rose, 1977). On Earth, basaltic systems may degas significant amounts of SO 2 (Carn et al., 2016) allowing reactions to occur with basaltic glasses that form rapidly on the flow surface or in small fountaining events. Basaltic fissure eruptions in the 2014-2015 Holuhraun eruption (Iceland) showed rapid formation of sulfate aerosols (Ilyinskaya et al., 2017). Although the source of the cations in the sulfate aerosols is unknown, a large proportion may be from basalt particles. Sulfates formed by reactions with SO 2 are most commonly observed in volcanic ash eruptions with anhydrite the most common phase produced (Ayris et al., 2013;Barone et al., 2016). Sulfates on volcanic ash particles from the Eyjafjallajökull (Iceland) eruption in 2010, include epsomite (MgSO 4 ·7H 2 O), melanterite (FeSO 4 ), hydrated thenardite (Na 2 SO 4 ·10H 2 O), and minor arcanite (K 2 SO 4 ; Gislason et al., 2011). Studies of Hawaiian basalts have shown that SO 2 -rich gases form an acid fog environment, altering basalts to form sulfate-rich coatings including anhydrite, gypsum, jarosite, natroalunite, and anatase, on amorphous silica rims (McCanta et al., 2014;Schiffman et al., 2006).
Although sulfate products from the reactions between volcanic gas and natural volcanic glass are observed in natural samples, the experimental data to constrain the mechanism of reaction were missing until relatively recently. McCollom et al. (2013) conducted experiments reacting a wet SO 2 -bearing vapor with basalt at 145°C for up to 137 days, observing the formation of coatings containing anhydrite, Fe-rich natroalunite, traces of other sulfates, and iron oxides on a layer of amorphous silica. Ayris et al. (2013) investigated reactions between dry SO 2 -rich gas and tephrite, phonolite, dacite, and rhyolite glasses at temperatures up to 800°C. They observed coatings rich in CaSO 4 form on the ash particles at rapid rates (within minutes) at temperatures of 300°C and above. They suggest that the Ca-availability in the glass drives the sulfateforming reaction. Renggli et al. (2019) studied dry, pure SO 2 reactions with anorthite-diopside (AnDi) composition glass starting materials and found a temperature dependence of the reaction products. At 600°C (below the glass transition temperature, T g ), CaSO 4 and MgSO 4 coatings formed on the An 15 Di 85 , An 36 Di 64 (eutectic glass), and An 48 Di 52 glasses. At 800°C, only CaSO 4 formed on all glasses (Renggli et al., 2019). In the presence of Ca in the glass, even at concentrations as low as 5 mol. %, CaSO 4 is the dominant reaction product on glasses in the anorthite-diopside-albite system (Renggli et al., 2019). Ayris et al. (2013) and Renggli et al. (2019) both show that the mobility of the sulfate-forming cation (e.g., Ca 2+ ) in the glass and its diffusion to the glass surface is the rate-limiting step in the reaction of SO 2 with glass starting material. These studies show that glass composition and structure are important controls for the reaction rate and sulfate compositions.
In sum, the experimental work shows that sulfur dioxide (gas) rapidly reacts with geologic materials via surface-mediated reactions, where SO 2 disproportionates to form sulfates and reduced sulfur (gas or sulfides; Burnham, 1979;Ayris et al., 2013;Henley et al., 2015Henley et al., , 2017Delmelle et al., 2018;King et al., 2018;Palm et al., 2018;Zolotov, 2018;Renggli et al., 2019). The reaction follows the following form: where M refers to a cation such as an alkali or alkaline earth cation (e.g., Ca 2+ , Mg 2+ , Na + ). A complete reaction with excess SO 2 results in the consumption of all alkali or alkaline earth cations. Depending on the phase and composition of the reacted MO silicate , the product silicate (+oxide) consists of SiO 2 and aluminosilicates.
The reaction between pure SO 2 and Fe-bearing basalts affects the valence state of multivalent elements, such as Fe 2+/3+ , in the reaction products because the gas has an effective oxygen fugacity (logfO 2 /1 bar, henceforth referred to as logfO 2 ) of logfO 2 = −12.1 at 700°C  i.e., 4.2 logfO 2 units above the Ni-NiO (NNO) buffer). The starting glass is efficiently oxidized in a H-free system by the inward diffusion of electron holes, which are charge compensated by the outward diffusion of divalent cations such as Ca 2+ and Mg 2+ , where diffusivities of these cations are the rate-limiting factors (Cook & Cooper, 2000;Cooper, 2020;Cooper et al., 1996). The oxidation process results in the concentration of CaO and MgO at the glass surface and the nucleation of Fe 2 O 3 in the oxidized glass (Cook & Cooper, 2000;Cooper et al., 1996).
These studies suggest that the fO 2 of the starting materials may have an important role in producing the reaction products because the initial oxidation state of Fe (Fe 3+ /Fe 2+ ) has a role in the formation of oxides (reaction (1)) like Fe 2 O 3 . This is discussed in more detail in section 4.1.
Here we investigate the effect of different fO 2 states of basalt glasses on the reaction with pure SO 2 at high temperatures (>600°C). We show that the reaction products include a range of different sulfates, oxides, and silica coatings which form within hours. We discuss the effect of different fO 2 in S-O-bearing gases on the stabilities of reaction products based on thermodynamic calculations. Using the calculated fO 2 for Earth, Mars, Venus, Io, the Moon, and Mercury, we predict the solid reaction products on the surfaces of these planetary bodies.

Starting Materials
The experiments were run with a synthetic Fe-free basalt, a tholeiitic basalt, and an alkali basalt with the latter two compositions simulating common basalts on planetary bodies (Basaltic Volcanism Study Project, 1981). The Fe-bearing tholeiitic basalt glass is from the 1921 Pu'u O'o eruption at Kilauea, Hawai‛i (Holloway & Burnham, 1972) and the alkali basalt is from San Carlos, Arizona (King et al., 2000). The tholeiitic and alkali basalts were prepared as part of previous experimental studies and are completely crystalfree (Dufresne et al., 2015;King et al., 2000). The glass compositions were determined using an electron microprobe and Mössbauer spectroscopy (Table 1 by Dufresne et al., 2015). Additional experiments were conducted with Fe-free basalt glass to evaluate the effect of Fe, and this material was synthesized from chemical grade oxide powders. The powders were ground and mixed before melting at 1500°C for 10 min, quenched, reground, and remelted twice at 1500°C. The glass starting materials were analyzed using a scanning electron microscope (SEM) to ensure that they were homogeneous and crystal free. The glass chips were polished to ¼-μm grit and washed with methanol to ensure a flat and clean surface.
The tholeiitic basalts have glass transition temperatures (T g ) between 715 and 745°C, whereas the T g of the alkali basalt is~640°C (Ryan & Sammis, 1981). Consequently, some of the experiments were conducted below T g with glass sensuo stricto, whereas for experiments conducted above T g these samples were strictly "supercooled melts." In the general discussion we refer to all samples as basalt glasses and use the term supercooled melts only when directly referring to specific experiments on basaltic substrates conducted above T g .

Experimental Conditions
Polished glass chips with diameters of~2 mm were placed on a Pt-tray which was suspended in a vertical gas mixing furnace at 600, 700, or 800°C for 1 hr and 24 hr and exposed to a 50-sccm (ml/min) flow of SO 2 . Before and after each SO 2 experiment, the furnace was purged with Ar for approximately 2 min to allow thermal equilibration before exposure to SO 2 and to remove any atmospheric contamination and to minimize escape of SO 2 into the laboratory. The experimental run products (reacted glasses) were immediately placed in a desiccator to minimize hydration of the reaction products . Three aliquots of sample were run so that enough sample was available for a wide range of analytical methods (like Palm et al., 2018). Cross sections of glasses were obtained by embedding the run products in epoxy resin and cutting the mounds. The samples were cut using a dry diamond saw. The cross sections were polished with diamond paste using ¼-μm diamond grit without any water. During the preparation of cross sections, the coating was lost from some glass surfaces. This may have allowed polishing material and possible contaminants to become embedded between the sample and the embedding epoxy resin, and as such, Al-rich contamination during polishing was considered when interpreting the analyses. The polished samples were cleaned with ethanol and compressed air.

Analytical Methods
The experimental run products were investigated using scanning electron microscopy, Raman spectroscopy, and nanoSIMS element mapping. These analytical methods revealed the chemistry and mineralogy of the different phases in the coatings on the reacted basalts and the chemical modification in the reacted substrates Mernagh et al., 2018;Palm et al., 2018).
The mineral phases in the surface coatings were identified using a Renishaw 2000 inVia Reflex Raman Spectrometer equipped with a standard confocal microscope at the Research School of Physics and Engineering at the Australian National University. A Renishaw diode-pumped solid state laser provided 532 nm laser excitation. The laser was focused with a 100X objective lens, with a resulting laser power of 3 mW on the sample and a spot size of~1 μm. Single Raman spectra were obtained using a 1-s integration time with 10 accumulations. At these conditions the laser may sample the material to a depth of 4 μm, depending on the laser transparency of the analyzed material (Mernagh et al., 2018). The 520.7-cm −1 band from a pure Si standard was used to test the alignment of the spectrometer prior to the analysis of the coatings.
The coatings formed by the SO 2 -glass reactions on the basalt glasses were observed with Field Emission Scanning Electron Microscopy (FESEM), using a Hitachi 4300 SE/N Schottky Field Emission instrument at the Centre for Advanced Microscopy at the Australian National University. Backscattered electron (BSE) images were collected, allowing the distinction of the different phases in the coatings and their textural relationships. Due to the roughness and fine grain size of some phases in the coatings, quantitative compositional analysis is not possible. Instead, we used Energy Dispersive Spectroscopy (EDS) with an 80-mm 2 silicon drift detector to collect qualitative elemental maps from the same areas where the BSE images were recorded. To increase the efficiency of element mapping, we increased the beam current to 1,000 pA by widening the condenser lens and reduced the spectrum processing time by 20%.
Surface phases and modification features observable in cross sections of the reacted glass substrates were analyzed with nanoscale secondary ion mass spectrometry (nanoSIMS), using a CAMECA nanoSIMS 50L instrument at the Centre for Microscopy, Characterization and Analysis, at the University of Western Australia. The samples were mounted in resin, polished and gold coated. We used a Hyperion (H200) oxygen ion (O − ) source, providing a spot size of 100 nm with a beam current of 12.5 pA and an ion impact energy of
Since our experiments were carried out in the absence of water, we expect that the MgSO 4 .nH 2 O was initially anhydrous. Nonetheless, using the main Raman SO 4 2− stretching modes, we can identify three differently hydrated magnesium sulfate phases including MgSO 4 ·H 2 O (kieserite), MgSO 4 ·2H 2 O (sanderite), and MgSO 4 ·3H 2 O (trihydrate; Wang et al., 2006). Despite placing the reacted glass samples in a desiccator immediately after taking them out of the gas-mixing furnace, Raman analysis was performed under air and so we suspect that hydration of MgSO 4 occurred after the experiments. This underlines the difficulty of appropriately handling gas-solid reaction products Palm et al., 2018) and the need to use a N 2 or Ar purge during analysis.
Sodium is incorporated in two different sulfates, including Na 2 SO 4 and the mineral glauberite (Na 2 Ca(SO 4 ) 2 ; Figure 1). The polymorph of Na 2 SO 4 could not be identified unambiguously because of its poor crystallinity and occurrence in a mixture with Fe 2 O 3 ( Figure 1). Glauberite is observed only on the alkali basalt supercooled melt preequilibrated under highly reducing conditions (~NNO−5) and reacted with SO 2 at 800°C (Table 2). Glauberite is identified by an intense band at 1,002 cm −1 in the Raman spectrum, which is distinct from anhydrite (1,016 cm −1 ; Buzgar et al., 2009;Lopez-Reyes et al., 2014). The sulfate spectra collected using Raman tend to have a broad amorphous band, which suggests that the sulfate coatings are thinner than the penetration depth of the Raman laser, resulting in both the coating and the glass being probed simultaneously (Mernagh et al., 2018).
Oxide phases include hematite and magnetite (Zoppi et al., 2008), anatase (Tompsett et al., 1995) and a Fe-Ti-(Al)-oxide. Hematite occurs in the coatings of all experiments, as both large crystals with diameters of up to 10 μm, but also as very fine grained material intermixed with Na 2 SO 4 , such as on the basalt glass Note. For each experiment, the phases in the surface coatings were determined using Raman spectroscopy and EDS (Figure 1), and the crystalline phases in the reacted substrates were inferred from nanoSIMS chemical maps (Figures 3-8).

Textures and Chemistry of the Coatings
The SEM images show that the coatings are discontinuously mounded on the top of the glass substrates. Gaps in the coatings are evident in areas where high Si-and Al-counts are measured; these elements are immobile and remain at the surface of the substrate during reaction with SO 2 (Figures S2-S8 in the supporting information).
The textures of the surface coatings fall into two categories ( Figure 2). First, fine-grained and equigranular coatings are found on the Fe-free basalt, the reduced alkali basalt preequilibrated at NNO−5 and the reduced tholeiitic basalt (NNO−1.5). Second, large anhydrite crystals, with lengths extending up to 100 μm (Figure 2), occur in the coatings on the more oxidized tholeiitic basalts (NNO and NNO+1) and the oxidized alkali basalt (NNO). Here we discuss the two types of textures separately.
On the Fe-free basaltic glass and the most reduced tholeiitic basalts (NNO −1.5), fine-grained CaSO 4 and MgSO 4 can be differentiated by their crystal habits ( Figure 2) and the EDS maps of S, Ca, and Mg ( Figures S2-S8). After reaction for 1 hr, the crystal habit of the CaSO 4 is poorly developed (Figures 2a and 2g), but as the grains mature with longer experimental durations (i.e., 24 hr), they establish facets (Fe-free basalt, Figure 2j) and grain-boundary triple-junctions, with dihedral angles of 120°(NNO−1.5, Figure 2d). The MgSO 4 grains are rounded, likely because the grains have taken up variable amounts of water after the experiments. In the BSE images ( Figure 2) the MgSO 4 grains can further be identified by their darker gray shade. The grain sizes of the coatings on the Fe-free basalt and the reduced NNO−1.5 basalt are similar after 1 hr and 24 hr ( Figure 2). The coatings on both of the latter glass substrates do not contain any Na 2 SO 4 (Figures S1-S4).
The textures of the coatings on the SO 2 -reacted reduced (~NNO−5) alkali basalt are similar to the fine-grained texture observed on the Fe-free basaltic glass, but with additional fine-grained glauberite and rounded hematite and magnetite crystals up to 20 μm in diameter. The sulfates are generally fine grained and the coating can be wrinkled and buckled in areas where the coating is thickest, due to lateral expansion of the coating (Renggli et al., 2019;. The coatings on the more oxidized (NNO and NNO+1) basalts are dominated by large elongated CaSO 4 crystals with grain sizes of up to 100 μm in length (Figure 2). For the basalts preequilibrated at the NNO-buffer, the grain sizes are consistently larger on the reacted tholeiitic basalt compared with the alkali basalt ( Figure 2). The coating on the surface of the reacted alkali basalt (NNO) is also more discontinuous (patchy) in comparison to the tholeiitic basalt.
The large CaSO 4 grains show a perfect cleavage perpendicular to the longest crystal axis (Figure 2). This cleavage plane corresponds to the (001) plane (Tröger, 2017), and hence the longest dimension of the elongated grains is the crystallographic c axis. Anhydrite has the largest thermal expansion coefficient along the crystallographic c axis, with a relative expansion of 2% from room temperature to 700°C (Evans, Sodium sulfate forms fine needles which are commonly twinned and have grain sizes of less than 2 μm (Figure 2). On the reacted alkali and tholeiitic basalts (preequilibrated at the NNO-buffer) Na 2 SO 4 is mainly observed in-between the CaSO 4 crystals and to a lesser degree associated with MgSO 4 ( Figure S6). On the most oxidized tholeiitic glass (NNO+1) fine-grained Na 2 SO 4 occurs in dense aggregates and is mixed with discrete, fine-grained Fe 2 O 3 crystals (Figures 2f and S8), that are revealed by Raman spectra (Figure 1).
Hematite and anatase are clearly observed by their light gray hues in the BSE images (Figure 2), high iron and titanium counts in the EDS maps (Figuress S1-S8) and Raman spectra (Figure 1). The grain size of hematite ranges from <1 μm on the basalt glass preequilibrated at NNO−1.5 to 20 μm on the NNO glass ( Figure 2). On the most oxidized glass (NNO+1) submicron Fe 2 O 3 grains are associated with Na 2 SO 4 aggregates (Figures 1, 2f, and S8). Anatase is less abundant than hematite, and grain sizes range from <1 to 10 μm.
In summary, the coating textures and mineral assemblages on the basalt glasses differ between (1) the Fe-free glass, the reduced (~NNO−5) alkali basalt and the most reduced tholeiitic basalt glass (NNO−1.5) and (2) the more oxidized tholeiitic basalt glasses (NNO and NNO+1). The assemblages are summarized in Table 2. In the Fe-free and reduced basalts (1), the coatings consist of equigranular grains of CaSO 4 and MgSO 4 , with no significant textural differences observed between 1-hr-and 24-hr-long experiments. In the oxidized basalts (2), the coatings are dominated by large CaSO 4 crystals, which increase in size in the 24-hr experiments  (Table 2) fill in the space between and occasionally overgrowing onto the large CaSO 4 crystals. This suggests that Na 2 SO 4 and MgSO 4 sulfates nucleate or grow, at least partially, after CaSO 4 . In the case of MgSO 4 , postexperiment hydration has likely affected our interpretation of its texture.

Chemistry of the Glass Substrate in Cross Section
Chemical element maps of cross sections measured using nanoSIMS reveal crystallites and chemical zones in the reacted basalts (Figures 3-7). The nanoSIMS maps show relative increases and decreases in count rates, which correlate with the relative concentrations of the elements. The figures are annotated to highlight features described in the text. Figure 8 shows profiles across the nanoSIMS maps, highlighting relative enrichments and depletions in the major elements from the surface to the modified basalt glass substrates.
In some cross sections of the reacted samples, the coating is poorly preserved. In these cases, the coating was lost when the cross sections were cut, polished, and embedded in epoxy. For example, the coating on the Fefree basalt glass is mostly lost (Figure 3), whereas it is well-preserved on the most oxidized (NNO+1) tholeiitic basalt glass ( Figure 6). In this case, the nanoSIMS map ( Figure 6) shows a sulfate coating on the glass surface with a thickness of~8 μm, predominantly consisting of CaSO 4 with some Na-rich sulfate. Iron oxides, Ti-oxides, and Fe-Ti-oxides are also observed in the coating layer (Figure 6). At depth, with a distance of 10-50 μm from the surface depending on the sample, the glasses remain homogeneous, unreacted, and crystal free.
The Fe-free basaltic glass substrate shows the least amount of modification after reaction with SO 2 for 24 hr at 700°C (Figure 3). Calcium and Mg are depleted at the glass surface (Figures 3a, 3b, and 8d). The concentration of Ca increases into the glass substrate over a distance of 5 μm, whereas Mg increases occur over a shorter distance of~2 μm. Sodium is enriched at the surface to a depth of 2-5 μm (Figure 8d). The depletion of Ca and Mg in the glass substrate correlates with the observation of CaSO 4 and MgSO 4 in the sulfate coating, whereas Na 2 SO 4 is not observed and hence Na is retained in the near-surface glass substrate. There is no change in Si, Al, and Ti in the glass substrate (Figures 3d-3f). No sharp peaks are observed in the major element profiles (Figure 8d), which suggests that no crystallites have nucleated on The tholeiitic glass preequilibrated at NNO−1.5 is modified in a similar manner to the Fe-free basalt glass ( Figure 4). Calcium is depleted at the surface of the glass substrate to a depth of~3 μm ( Figure 4a) and Na is enriched in the same zone at the glass surface (Figures 4c and 8a). Magnesium-rich crystals are present in the reacted basaltic substrate to a depth of 20 μm (Figure 4b), but the highest density zone of Mg-rich crystallites occurs within 3 μm of the surface. The largest crystallites have diameters of up to 2 μm. Silicon-and Fe-bearing crystallites form in the glass substrate within 1-3 μm of the surface (Figures 4d and 4f).

Journal of Geophysical Research: Planets
In the reacted tholeiitic glass preequilibrated at NNO (Figure 5), Ca is depleted at the surface to a depth of <10 μm (Figure 5a). In comparison, Na becomes increasingly enriched toward the glass surface from a depth of 20 μm (Figure 5c). Nucleation of Fe-, Na-, and Mg-rich spherulites is observed throughout the sample, with decreasing numbers and increasing grain sizes from the surface toward the interior of the glass (Figures 5b, 5c, 5f, and 8b). The crystals have Fe-rich cores and Na-concentrations are elevated throughout the spherulites. This is evident when comparing Figures 5c and 5f, where high Na and Fe count rates are observed at a depth of more than 15 μm from the surface of the reacted glass. We interpret the larger spherulites with diameters of up to 3 μm as pyroxenes (e.g., NaFe 3+ Si 2 O 6 , aegirine). They are surrounded by Mgrich material, which may correspond to MgSiO 3 (enstatite or clinoenstatite). The Mg-rich rims become

Journal of Geophysical Research: Planets
The modification of the most oxidized tholeiitic basalt glass (NNO+1) differs from the other glasses in its texture and the composition of the nucleated phases ( Figure 6). Calcium is strongly depleted at the surface to a depth of 12 μm (Figures 6a and 8c). Mushroom-shaped features, defined by stong gradients in Mg, Na, and Fe, are observed up to 20 μm from the glass surface. These are highly enriched in Na and Fe and include a number of different crystallites (Figures 6c and 6f). Sodium is enriched throughout the modified zone, whereas Fe is depleted within 1.5 μm of the glass surface and enriched further into the glass (Figures 6c  and 6f). Magnesium is enriched around the mushroom-shaped modification features but is also depleted within~2 μm from the surface (Figure 6b). Crystals are evident as peaks in the nanoSIMS line profile (Figure 8c), similar to the NNO glass (Figure 8b), and include Mg-rich spherulites surrounded by Fe-and Na-rich material, Fe-rich crystallites ( Figure 6f) and SiO 2 needles (Figure 6d), and Al-rich material (Figure 6e), possibly Al 2 O 3 . The different nanoSIMS figures (Figures 3-7) show that the Al-rich material fills the gap between reacted glass substrate, coating phases, and the surrounding epoxy resin. This textural setting could be explained by it being derived from either contamination or as a reaction product. Thermodynamic calculations suggest that Al 2 O 3 could form as a reaction product, and it has been observed on the surface of pyroxene reacted with SO 2 . We sought to determine if the Al 2 O 3 could be detected on the surface of an unpolished sample by Raman spectroscopy or scanning electron microscopy, rather than just in the polished section. The surface analyses did not reveal the Al 2 O 3 phase, and this might be due to it either not being present or alternately, below the detection limits of the techniques since nanoSIMS indicates that it is found in very thin layers. At this time, we are not able to identify the source of the Al-enrichment as being either contamination or a reaction product.
The nanoSIMS maps of the reduced alkali basalt supercooled-melt (~NNO−5) reacted with SO 2 at 800°C for 1 hr show the presence of Ca and Na in the coating, which agrees with the observation of the mineral glauberite (Na 2 Ca(SO 4 ) 2 ; Figures 1 and 7). The basalt is modified to a depth greater than 50 μm. We propose that modification and crystallization results from both reaction with SO 2 and from heating the silicate glass above its glass transition temperature. Near-surface crystallization of supercooled melts reacting with SO 2 was previously observed in experiments in the anorthite-diopside system (Renggli et al., 2019). The modified alkali basalt shows a stratification of major element enrichment (Figures 7 and 8e). This is best observed for Ca, Mg, and Na, which are the most mobile elements in this reaction, as observed in their enrichment in the coating (Figure 2). Crystallization products in this basalt were previously identified as orthopyroxene (hypersthene; Palm et al., 2018).
In summary, the observed modification of the basalt glasses after 24 hr of reaction with SO 2 extends into the glass substrate to a depth of less than 5 μm from the surface in the case of the Fe-free and the NNO−1.5 glasses (Figures 8d and 8e). The more oxidized glasses are modified to a depth of~20 μm (NNO+1) to more than 50 μm (NNO) from the surface (Figures 8b and 8c). The alkali basalt supercooled melt preequilibrated at NNO−5 is modified to a greater depth due to the reaction above T g , which allows the rapid crystallization during the experiment. The formation of sulfates on the glasses is reflected by a strong depletion of Ca close to the surface of the glass substrate. Magnesium and Fe are also depleted near the glass surface, whereas Na is enriched near the surface of the glass substrate in the tholeiite glass experiments. Crystallites in the modified Fe-bearing basalt glasses are rich in Mg and Fe, with some Na incorporated in the Fe-rich crystallites. We were not able to directly identify the minerals, but the nanoSIMS maps suggest that they are NaFe 3+ Si 2 O 6 (aegirine) and MgSiO 3 (enstatite or clinoenstatite). Hypersthene was previously identified in alkali basalt glass that reacted with SO 2 . The reaction products in the glasses were caused by the reaction with SO 2 at high temperatures. The unreacted materials were homogenous and crystal free (e.g., Figures 3-8 show homogeneous composition at depth). Upon annealing of the glasses at 800°C in air we did not observe the chemical alteration textures observed in the SO 2 reaction experiments (supporting information).

Discussion
We argue that the formation of phases in the coating is controlled by the role of the cations (Ca 2+ , Mg 2+ , Na + , Fe 2+/3+ ) in the glass structure and their ability to migrate to the glass surface (mobility). Calcium diffusion has previously been suggested as the rate-limiting factor in the reaction between SO 2 -bearing gas and glassy ash, which forms CaSO 4 (Ayris et al., 2013; Delmelle et al., 2018). The mineralogy of the crystallites in the substrate is controlled by the locally changing composition, which is due to the loss of Ca, Mg, Na, and Fe to the coating. First, we discuss the basaltic glass structure and how it may affect the reaction mechanism. We then discuss how the reaction oxidizes the Fe-bearing basalt glasses and supercooled melts, resulting in an increase in Fe 3+ /Fe tot , and present a model for the reaction mechanism. We present thermodynamic calculations showing the predominance of Fe, Na, Ca, and Mg phases, including oxides, sulfides, and sulfates, as a function of logfO 2 and logfSO 2 in the gas phase and relate our experiments to gas-solid processes that occur in natural systems on planetary bodies including Earth, Venus, Mars, Io, the Moon, and Mercury.

Role of the Glass Structure and Redox State on Reaction Products
Glass structure and composition have been shown to be important controls on the reaction with SO 2 (Ayris et al., 2013;Renggli et al., 2019;. In the structure of a silicate glass, Si 4+ , Ti 4+ , Fe 3+ , and Al 3+ cations are tetrahedrally (T)-coordinated by O 2− ions. These tetrahedral building blocks form a network with short-to medium-range order, but no long-range order (Mysen & Richet, 2005). The O 2− ions may either bond to two network forming T-cations, forming T-O-T bonds and acting as bridging oxygens (BO), or they may link a T-cation with a network modifying cation (M; e.g., Na + , K + , Mg 2+ , Ca 2+ , and Fe 2+ ), forming T-O-M bonds and acting as nonbridging oxygens (NBO). The number of nonbridging oxygens per tetrahedrally coordinated cation (NBO/T) describes the degree of polymerization of the amorphous network (Mysen & Richet, 2005). Tetrahedrally coordinated Al 3+ and Fe 3+ cations further require charge balancing by a monovalent or divalent cation as they are negatively charged ([M 3+ O 4 ] -). Alkali earths in particular may change their role in the glass structure and act as charge compensators, instead of network modifiers (Angeli et al., 2000;Le Losq et al., 2014;Le Losq & Neuville, 2017;Neuville & Mysen, 1996).
In more reduced glasses (~NNO−5, NNO−1.5, and NNO) the Na + concentration is sufficient to charge balance all tetrahedrally coordinated Fe 3+ cations (Fe 3+ /Na + < 1; Table 1). However in oxidized glasses, alkalis are not sufficient to completely charge balance all of the tetrahedrally coordinated trivalent cations in the glass, and some divalent cations (Ca 2+ , Mg 2+ , Mn 2+ ) are required for charge compensation. The need for charge compensation increases in the more oxidized glasses (NNO+1), where the abundance of tetrahedrally coordinated Fe 3+ is higher (Fe 3+ /Fe tot = 0.79 for the tholeiitic basaltic glass preequilibrated at NNO+1).
The reaction of SO 2 with Fe-bearing basaltic glass is an example of a surface-mediated reaction that includes oxidation . Because sulfur is a polyvalent element (S 2− , S − , S 0 , S 4+ , S 6+ ), the redox state of both the gas and solid reactants play a role in the reaction. Here, we assume that the SO 2 used in the experiments is pure and dissociates at equilibrium to form other gas species including S 2 , S 3 , SO 3 , SO, S 2 O, and O 2 Zolotov & Fegley, 1999). Using a Gibbs Free Energy minimization calculation of the equilibrium dissociation of SO 2 , the fO 2 of the gas is logfO 2 = −12.1 at 700°C and 1 bar, equivalent to 4.2 log units above the NNO-buffer (O'Neill & Pownceby, 1993). Consequently, the SO 2 reaction with the more reduced basalts (NNO−5 to +1, Table 2) imposes a redox gradient.
In a H-free system where a silicate glass reacts with an oxidizing gas, this electrochemical potential gradient is most efficiently dissipated by the rapid diffusion of electron holes into the glass to oxidize Fe 2+ to Fe 3+ . The inward diffusion of the electron holes is charge compensated by the counterdiffusion of Ca and Mg. This mechanism is explored by Cooper and coworkers (Cook & Cooper, 2000;Cooper, 2020;Cooper et al., 1996). They show that multivalent element-bearing aluminosilicate glasses act as semiconductors for electron holes, allowing their rapid diffusion. In this oxidation mechanism, the charge-compensating diffusion of Ca 2+ and Mg 2+ is the rate-limiting step. At the glass surface, Ca 2+ and Mg 2+ form MgO and CaO coatings in such an oxidation reaction (Cook & Cooper, 2000;Cooper, 2020;Cooper et al., 1996). Finally, the oxidized Fe 3+ atoms require charge compensation if they are tetrahedrally coordinated, which results in the development of a Na-front at the oxidation front in the glass (Burkhard, 2001;Cook et al., 1990;Cook & Cooper, 2000). We observe a similar Na-front in the nanoSIMS maps and profiles obtained from our experiments (Figures 3-8), suggesting that this oxidation mechanism operates in our experiments. The nanoSIMS maps of a cross section of the reacted tholeiitic basalt (e.g., preequilibrated at NNO+1; Figure 6) show correlations between Fe-enrichment and Na-enrichment in the near-surface glass substrate. In these areas, Fe is likely present as Fe 3+ and charge compensated by Na + .

Journal of Geophysical Research: Planets
We argue that the new minerals observed both on the surface and within the basalt glasses reacted with SO 2 (Figures 3-8) are reaction products of the gas-solid reaction. To demonstrate the cause of the alteration is not the exposure to high temperatures we conducted an annealing experiment at 800°C in air, without SO 2 . A tholeiitic glass preequilibrated at NNO was kept at 800°C in air for 24 hr ( Figure S1). In cross section, the annealed glass shows no changes from the original material: it is homogeneous and crystal free. At the surface we observed micron-sized crystals of CaO and MgO, which is in agreement with previous studies investigating the oxidation of basaltic glasses (Cook & Cooper, 2000;Cooper, 2020;Cooper et al., 1996). Our experiments do not show any evidence for CaO and MgO, instead Ca and Mg sulfates. Both experiments are evidence for the diffusion of Ca and Mg in the glass toward the surface during annealing (as described by .

Sulfate Formation
The sulfate-forming cations Ca, Mg, and Na are mobilized and concentrated in surface coatings by two different mechanisms. First, as discussed above, the basalt oxidation mechanisms first proposed by Cook et al. (1990) results in the formation of CaO and MgO coatings on the glass surface. In the presence of SO 2 these alkaline earth oxides react at the surface of the glass to form sulfates (see equation (1)). Second, the chemisorption reaction of SO 2 with the glass surface imposes a chemical potential gradient from the surface to the interior of the glass and mobilizes the sulfate-forming cations to the surface. Both mechanisms are linked and cannot be separated, but the overall rates are limited by the diffusivities of the cations . As the near-surface glass becomes increasingly depleted in the sulfate-forming cations the diffusivities change and the overall reaction rate does not remain constant. Furthermore, the changes in the glass composition result in the formation of SiO 2 and other crystallites that can armor the surface of the reactant, as observed in our experiments (Figures 3-8). This crystallization below the glass surface further limits the diffusion of the sulfate-forming cations to the surface and reduces the overall reaction rate.
Iron present at the surface of the glass is oxidized by SO 2 to form Fe-oxide minerals (e.g., hematite and magnetite). A titanium-rich oxide mineral, anatase (TiO 2 ), was also identified as a reaction product present in the coatings formed on the Fe-free and tholeiitic basalts reacted with SO 2 . This is evidence that Fe and Ti may diffuse to the surface of the coating as the reaction progresses. The mineralogy of the Fe-oxide minerals formed is dependent on the temperature and gas partial pressure-particularly fO 2 -of the experiment Figure 9); magnetite is only present in the 800°C experiments on Fe-bearing basalts.
Thermodynamic calculations of the reaction between SO 2 and silicate minerals produce equilibrium assemblages of sulfates, aluminosilicate phase(s), metal oxides, and S 2 gas . In nature, complete reaction and attainment of equilibrium between minerals or glasses and SO 2 is unlikely and only occurs locally. For example, in the subvolcanic environment, sulfate veins form via the reaction of SO 2 with volcanic rocks over length scales of millimeters to meters (Casas et al., 2019;Henley et al., 2015Henley et al., , 2017Henley & Seward, 2018). More extensive reaction is limited by mass transport of cations in the reacted solids, transport of SO 2 through veins and pore space, overall availability of SO 2 to the reaction, transport of S 2 in the system, and the lifetime of the volcanic system. In explosive volcanic eruptions the reaction is limited by the short lifetimes of high-temperature plumes of seconds to few minutes (Ayris et al., 2013), even though SO 2 is available in excess and the surface area of the fine-grained volcanic ash is large and favorable for the reaction.
Similar to nature, it was not possible to attain equilibrium conditions in the experiments within reasonable timescales. Instead, our observations show a snapshot of the reaction products on the way to equilibrium and represent what we expect in nature. Even in the coating at the surface equilibrium is not attained. The gassolid reaction occurs in an open system at the glass surface, where SO 2 is continuously transported to the surface and reduced S-gas is removed (equation (1)). Similarly, the sulfate-and oxide-forming elements Ca, Mg, Na, Fe, and Ti diffuse to the reaction interface at changing rates throughout the experiment.

Implications for the Inner Planets, the Moon, and Io
In our experiments we varied the redox state of the basalt glass starting materials, but we did not vary the composition of the reacting gas phase. Here, we use thermodynamic calculations to address the effect of different fO 2 and fSO 2 on the stabilities of sulfate, sulfide, and oxide compounds of Fe, Na, Ca, and Mg. Figure 9 shows predominance plots of the prevailing phases in the systems Fe-S-O, Na-S-O, Ca-S-O, and Mg-S-O at 700°C and atmospheric pressure, based on minimum Gibbs energies calculated using the program HSC9 by Outotec (Roine, 2015). The theoretical equilibrium condition in our experiments falls in the predominance fields of Fe 2 O 3 , Na 2 SO 4 , CaSO 4 , and MgSO 4 (Figure 9), confirming the stability of the phases formed in the experiments.
The oxygen fugacity of terrestrial planetary bodies encompass a range of almost 30 orders of magnitude at their surfaces. We overlay estimated present and past SO 2 and O 2 fugacities of magmatic systems and atmospheric compositions over the phase predominance plots (Figure 9). This allows us to make predictions on the stability of sulfate, sulfide, and oxide phases of Fe, Na, Ca, and Mg on the surfaces of these planetary bodies, where reactions with S-bearing gases occur, assuming 1 bar of pressure.
highest observed SO 2 abundance in a volcanic gas on Earth is 0.21 mole fractions are from the Kilauea lava lake (Gerlach, 1980). As meteoric water is entrained in volcanic gas prior to eruption, H 2 S becomes more abundant (Henley & Seward, 2018) and SO 2 abundances may drop below 10 −4 mole fractions.
Magmas on Mars are more reducing than on Earth with fO 2 estimates ranging from slightly below NNO−5 to NNO (Castle & Herd, 2017;Gaillard et al., 2013;Gaillard & Scaillet, 2009;Hirschmann & Withers, 2008). Yet the atmosphere is oxidizing (logfO 2~− 5) resulting in abundant oxidized hematite (King & McSween, 2005;Morris et al., 1989), suggesting that redox conditions in gas-solid reactions on Mars's surface could have been more oxidizing. Gaillard and Scaillet (2009) suggested SO 2 abundances of up to 0.1 mole fractions in a martian volcanic gas. Lower limits may be significantly below 10 −4 mole fractions in an early Martian volcanic gas (Gaillard & Scaillet, 2009). Martian conditions fall in the stability fields of sulfates and oxides, but Na-, Ca-, and Mg-sulfides could occur in the lower range of redox conditions suggested for magmatic processes (Figure 9). Implications of these and other experiments for the fine material on Mars is the topic of ongoing work.
The oxygen fugacity of primary basalts on Venus is unknown and its atmosphere ranges from~NNO+1 tõ NNO+9 (Fegley et al., 1997;Vandaele et al., 2017) which falls in the range of conditions estimated for Earth and Mars (Figure 9). The modern atmosphere of Venus contains~150 ppm SO 2 and S-rich deposits on the surface likely formed via reactions between SO 2 and the basaltic surface (Zolotov, 2018, and references therein). The high S-abundance on the surface suggests that concentrations in volcanic gases and the atmosphere may have exceeded modern concentrations in the past (Zolotov, 2018). Zolotov (2018) discusses the stabilities of different rock-forming silicates, sulfates, sulfides, and carbonates at modern pressure and temperature conditions on Venus in more detail, suggesting that pyrite may be stable and that anhydrite is the only stable sulfate on the surface. For example, due to the presence of C-bearing species Mg forms carbonates instead of sulfates and the most stable Na-compound is albite (Zolotov, 2018).
Volcanic plumes on Io are S-rich and primarily consist of SO 2 and S 2 with oxidation states of NNO−3.3 to NNO+4 (Zolotov & Fegley, 1999. These conditions span the stability fields of sulfates and sulfides (Figure 9). At the more reducing S 2 -rich conditions (<NNO) Fe, Na, Ca, and Mg may be stable as sulfides on Io.
The Moon and Mercury are significantly more reducing than the other planetary bodies discussed here, with redox conditions below the iron-wüstite buffer (~NNO−5.3; Rutherford & Papale, 2009;Kerber et al., 2009 ;McCubbin et al., 2012 ;Zolotov, 2011). Due to the lack of atmospheres and modern volcanic gases on both bodies, estimates for SO 2 concentrations are more difficult. Lunar volcanic gas models suggest that SO 2 may have occurred at up to 10 −3 mole fractions (Renggli et al., 2017). Reactions between such a S-rich gas and the lunar regolith may have contributed to the formation of volatile-rich rocks such as the Apollo 16 "rusty rock," which contains sulfides and chlorides (Day et al., 2017;Shearer et al., 2014). In addition to Fesulfides, Na-, Ca-, and Mg-sulfides may also be stable at these conditions ( Figure 9). Mercury is assumed to be more reducing than the Moon with conditions of NNO−8 to NNO−11 (Kerber et al., 2009;McCubbin et al., 2012;Zolotov, 2011). High S-abundances were observed on Mercury (Nittler et al., 2011;Starr et al., 2012;Weider et al., 2012). In Figure 9 we use logfSO 2 estimates of the Moon for Mercury, and sulfur-bearing volcanic gas species have likely contributed to eruptive processes on Mercury (Kerber et al., 2011;Zolotov, 2011;Zolotov et al., 2013). Oldhamite (CaS) and niningerite (MgS) have been suggested as major sulfide species on Mercury's surface (Nittler et al., 2011). This is in agreement with our predictions as shown in Figure 9.
Finally, the availability of SO 2 in a natural volcanic gas is not only controlled by the bulk abundance of S in the gas, but also by the oxygen fugacity of the gas. With increasingly reducing conditions SO 2 may be reduced to SO, S 2 O, and S 2 . In reduced C-bearing gases, such as predicted for the Moon and Mercury, reduced S-gases may include CS 2 , COS, and CS (Renggli et al., 2017;Zolotov, 2011). In water-rich volcanic gas such as on Earth, the dominant reduced S-bearing gas species is H 2 S (Henley & Seward, 2018). These reduced S-gas species have different molecular properties and their behavior during chemisorption reactions differs from SO 2 . This difference may particularly affect the kinetics of the gas-solid reaction, whereas the reaction may still result in the formation of sulfides as predicted in Figure 9. Our experiments using pure SO 2 do not allow predictions on the kinetics of reactions with other S-gas species as they occur in natural volcanic 10.1029/2019JE006045 Journal of Geophysical Research: Planets gases. Specifically, future work will require experiments exploring gas-solid reactions with S at reducing conditions applicable to the Moon or Mercury.

Conclusions
Our experiments reveal that reactions between SO 2 and basaltic glasses produce CaSO 4 , MgSO 4 , Na 2 SO 4 , Fe-, Fe-Ti-(Al)-, and Ti-oxides, and silica coatings with pyroxenes and Fe-Ti-oxide crystallites in the glass substrates. This is in contrast to previous experimental studies, which observed a prevalence of CaSO 4 as the sole or dominant reaction product in reactions between SO 2 and silicate glasses (Ayris et al., 2013;Renggli et al., 2019). The experiments reveal a complex gas-solid reaction mechanism, whereby mobile elements such as Ca, Mg, Na, and Fe are lost to surface coatings rich in sulfate and oxide minerals, and silicate phases nucleate close to the surface of the glass. Our characterization of the products of SO 2 -basalt experiments may be applied to understanding the mineralogy and chemistry of planetary settings where high-temperature gas-solid reactions occur, such as volcanic environments where SO 2 -bearing gas and basalts are observed to coexist.